Validation of a visual attention model in a driving field test: difficulties and benefits

HAL Id: hal-00873671

https://hal.archives-ouvertes.fr/hal-00873671

Submitted on 10 Dec 2020

HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

Validation of a visual attention model in a driving field test: difficulties and benefits

Sophie Lemonnier, Roland Bremond, Lara Desire, Thierry Baccino

To cite this version:

Sophie Lemonnier, Roland Bremond, Lara Desire, Thierry Baccino. Validation of a visual attention model in a driving field test: difficulties and benefits. IAAP School on Applied Cognitive Research, Apr 2013, Paris, France. 1p. �hal-00873671�

www. ifsttar ^.fr

Each step of the experiment protocol leads to some difficulties, but each difficulty introduces variability in the data, and thus enlarges the field of potential applications.

[email protected]

Validation of a visual attention model in a driving field test:

difficulties and benefits. Lemonnier, S. [1,2], Brémond, R. [2], Désiré, L. [3], and Baccino, T. [1]

[1]

Paris 8, LUTIN -

IFSTTAR, LEPSiS -

CETE-Ouest

The First IAAP School on Applied Cognitive Research, Paris 2013

References

Smilek, D., Birmingham, E.,Cameron, D., Bischof, W.F., & Kingstone, A., 2006. Cognitive ethology and exploring attention in real world scenes. Brain Research, 1080, 101-119.

Tatler, B., Hayhoe, M., Land, M. F., Ballard, D., 2011. Eye guidance in natural vision : Reinterpreting salience. Journal of Vision 11 (5), 1–23.

Wickens, C. D., Goh, J., Helleberg, J., Horrey, W. J., Talleur, D. A., 2003. Attentional models of

multitask pilot performance using advanced display technology. Human factors, 45 (3), 360-380.

Background

Method

Discussion

Variability intra- and inter- of participants and situations

Although test field experiments are costly, some advantages emerge. Among the benefits of generalizing a visual attention model to true driving situations, its applications to road safety, road design, driving assistance systems and traffic simulation become easier. In addition, such an experiment on the road produces a large database of behavioral data, ready for future analysis.

Top-down

task, experience, expectations, goal

Top-down

task, experience, expectations, goal

Bottom-up

salience, effort

Bottom-up

salience, effort

Expectancy

amount of information present in an area

Expectancy

amount of information present in an area

Value

relevance of

information for a task

Value

relevance of

information for a task

Road sign

give way / stop /priority

Road sign

give way / stop /priority

Traffic

0 – few – many driver

Traffic

0 – few – many driver

Eye-tracking pattern

fixation duration saccade amplitude

Eye-tracking pattern

fixation duration saccade amplitude

Visual search

exploration of the visual scene

Visual search

exploration of the visual scene

Input / Output Cognitive process

Material: a vehicle mounted eye-tracking system

(SmartEye).

Driving situation:

anticipation of a crossroads.

Intermediates variables = road sign [value] + traffic density [expectancy].

Results: data analysis in progress.

Model

Visual attention = Bottom-up (i.e. saliency) and Top-down processes (i.e. goal, task, expectancy; +++ in the literature).

The more complex the processes under study, the more necessary it is to have an ecological experimental setting.

 Driving is an interesting task to study top-down models of visual attention (Tatler et al., 2011).

Wickens et al. (2003) proposed a model of visual attention predicting in which areas of interest people get information. Visual attention depends on two TD parameters: expectancy and value of information. We are currently testing this model with driving field test.

These differences in terms of participants and crossroads environments help to generalize our results to many driving situations.

(Smilek et al., 2006) Itinerary

includes 5 crossroads per road sign (i.e. give way, stop, priority). Some detours were necessary because of the constraints of the road network, which resulted in a 2 hours trip per participant.

A benefit of this long trip is the fact that the participants cannot guess the hypothesis under study.

Traffic density

can only be controlled a posteriori, with some intra-modality variability (i.e. no, low, dense traffic).

Eye-tracking calibration is complicated, including variable duration, with

expected consequences in terms of the

participants’ state of mind.

Surprising events

But also:

the weather, the familiarity of the trip, interactions with other drivers, and more

… all these factors contribute to the

variability of the

encountered situations.